Anthropic is releasing Opus 4.5 with new integrations with Chrome and Excel

Anthropic on Monday announced the Opus 4.5, the latest version of its flagship model. They are the last models of the Anthropic 4.5 series to be released, following the launch of Sonnet 4.5 in September and Haiku 4.5 in October.

As expected, the new version of Opus has cutting-edge performance on a range of benchmarks, including coding benchmarks (SWE-Bench, Terminal-bench), tool usage (tau2-bench, MCP Atlas) and general problem solving (ARC-AGI 2, GPQA Diamond).

Notably, Opus 4.5 is the first model to score more than 80 percent on the SWE-Bench test, a respected encoding benchmark.

Anthropic also emphasized the computer use and spreadsheet capabilities of Opus, and launched a number of parallel products to showcase how the model holds up in those settings. Along with the Opus 4.5, Anthropic will be making it Cloud for Chrome and Claude for Excel The products – previously experimental – are more widely available. The Chrome extension will be available to all Max users, while the Excel-focused model will be available to Max, Team, and Enterprise users.

Opus 4.5 also comes with memory improvements for long-context operations, which require significant changes in how the model manages its memory.

“There are improvements we’ve made to the quality of overall long context in training with Opus 4.5, but context windows won’t be enough on their own,” Diane Na Bin, head of product management for research at Anthropic, told TechCrunch. “Knowing the right details to remember is really important as well as having a longer contextual window.”

These changes have also enabled the long-awaited “Endless Chat” feature for paid Claude users, which will allow chats to continue uninterrupted when the form reaches its context window. Instead, the form will compress its context memory without alerting the user.

Many of the upgrades are made with proxy use cases in mind, particularly scenarios where Opus acts as a master proxy driving a group of Haiku-powered sub-agents. Managing these tasks requires strong control of working memory, which is where the memory improvements described by Ben really show their value.

“This is where basics like memory become really important,” says Ben, “because Claude needs to be able to explore code bases and large documents, as well as know when to step back and re-examine something.”

Opus 4.5 will face stiff competition from other recently released flagship models, most notably OpenAI’s GPT 5.1 (released November 12) and Google’s Gemini 3 (released November 18).